Speech interpreter with a unified grammer compiler

Abstract

The present invention provides a unified grammar for a speech interpreter capable of real-time speech understanding for user applications running on a general purpose microprocessor-based computer. The speech interpreter includes a unified grammar (UG) compiler, a speech recognizer and a natural language (NL) processor. The UG compiler receives a common UG lexicon and unified grammar description, and generates harmonized speech recognition (SR) and NL grammars for the speech recognizer and natural language processor, respectively. The lexicon includes a plurality of UG word entries having predefined characteristics, i.e., features, while the UG description includes a plurality of complex UG rules which define grammatically allowable word sequences. The UG compiler converts the complex UG rules (complex UG rules include augmentations for constraining the UG rules) into permissible SR word sequences and SR simple rules (simple rules do not include any augmentation) for the SR grammar. The SR grammar is a compact representation of the SR word entries corresponding to the UG word entries, permissible SR word sequences and simple SR rules corresponding to the augmentations of the complex UG rules. The NL grammar provides the NL processor with NL patterns enabling the NL processor to extract the meaning of the validated word sequences passed from the speech recognizer.

What is claimed is: 
     
       1. A method for converting a lexicon and a unified grammar (UG) description into a speech recognition (SR) grammar for controlling a speech recognizer, the lexicon including a plurality of UG word entries having predefined characteristics and the UG description including a plurality of UG rules, each UG rule having at least one augmentation, the UG rules defining grammatically allowable word sequences, the method being implemented as a plurality of program instructions and data stored in a memory of a system including a central processing unit (CPU) with the memory coupled to the CPU, a system bus coupled to the CPU, and an input means coupled to the system bus for receiving a digital signal representative of an acoustic signal, the method including the steps of: generating from the lexicon a plurality of SR word entries in said SR grammar corresponding to said plurality of UG word entries;   generating from the lexicon and the UG description a plurality of permissible SR word sequences and a plurality of SR rules in said SR grammar, said SR word sequences and said SR rules corresponding to the augmentations of said UG rules; and   storing said SR word entries, said permissible SR word sequences and said SR rules in the memory.   
     
     
       2. The method of claim 1 wherein said step of generating said SR word sequences includes the step of generating at least one said permissible SR word sequence for each grammatically allowable word sequence defined by at least one said UG rule, and wherein at least one word of at least one of said permissible SR word sequences is selected from said UG word entries by matching a predefined characteristic of said selected word to an element of one of said UG rules. 
     
     
       3. The method of claim 1 wherein said step of generating said SR word sequences include the step of generating one said SR rule corresponding to at least two of said grammatically allowable word sequences defined by one of said UG rules. 
     
     
       4. The method of claim 1 wherein said speech recognizer comprises a second plurality of program instructions and data stored in the memory. 
     
     
       5. The method of claim 1 wherein said speech recognizer comprises a plurality of program instructions and data stored in a second memory of a second computer system. 
     
     
       6. The method of claim 1 further comprising the step of generating a plurality of derivative word forms in said SR grammar corresponding to said plurality of UG word entries. 
     
     
       7. The method of claim 1 further comprising the step of generating a natural language (NL) grammar for controlling an NL processor coupled to said speech recognizer. 
     
     
       8. The method of claim 7 further comprising the step of propagating the predefined characteristic of said selected word to the NL processor. 
     
     
       9. The method of claim 1 wherein an element of one of said plurality of SR rules refers to another one of said plurality of SR rules. 
     
     
       10. The method of claim 1 wherein the step of generating SR rules includes the step of converting the UG description into a LISP format. 
     
     
       11. The method of claim 1 wherein the step of generating a first SR rule includes the step of assigning a true value to a feature associated with the corresponding UG rule. 
     
     
       12. A method for recognizing a digitized acoustic signal with a speech recognizer controlled by a speech recognition (SR) grammar corresponding to a lexicon and a unified grammar (UG) description, the lexicon including a plurality of UG word entries having predefined characteristics and the UG description including a plurality of UG rules, each UG rule having at least one augmentation, the UG rules defining grammatically allowable word sequences, the method being implemented as a plurality of program instructions and data stored in a memory of a system including a central processing unit (CPU) with the memory coupled to the CPU, a system bus coupled to the CPU, and an input means coupled to the system bus for receiving the digitized acoustic signal, the method including the steps of: generating from the lexicon a plurality of SR word entries in said SR grammar corresponding to said plurality of UG word entries;   generating from the lexicon and the UG description a plurality of permissible SR word sequences and a plurality of SR rules in said SR grammar, said permissible SR word sequences and said SR rules corresponding to the augmentations of said UG rules; and   comparing the digitized acoustic signal with said SR word entries, said SR word sequences and said SR rules for generating a validated word sequence corresponding to the digitized acoustic signal for a natural language processor.   
     
     
       13. The method of claim 12 wherein said step of generating said SR word sequences include the step of generating at least one said permissible SR word sequence for each grammatically allowable word sequence defined by one of said UG rules, and wherein at least one word of at least one of said permissible SR word sequences is selected from said UG word entries by matching a predefined characteristic of said selected word to an element of one of said UG rules. 
     
     
       14. The method of claim 12 wherein said step of generating said SR word sequences include the step of generating one said SR rule corresponding to at least two of said grammatically allowable word sequences defined by one of said UG rules. 
     
     
       15. The method of claim 12 wherein said speech recognizer comprises a second plurality of program instructions and data stored in the memory. 
     
     
       16. The method of claim 12 wherein said speech recognizer comprises a plurality of program instructions and data stored in a second memory of a second computer system. 
     
     
       17. The method of claim 12 further comprising the step of generating a plurality of derivative word forms in said SR grammar corresponding to said plurality of UG word entries. 
     
     
       18. The method of claim 12 further comprising the step of generating a natural language (NL) grammar for controlling an NL processor coupled to said speech recognizer. 
     
     
       19. The method of claim 18 further comprising the step of propagating the predefined characteristic of said selected word to the NL processor. 
     
     
       20. The method of claim 12 wherein an element of one of said plurality of SR rules refers to another one of said plurality of SR rules. 
     
     
       21. The method of claim 12 wherein the step of generating SR rules includes the step of converting the UG description into a LISP format. 
     
     
       22. The method of claim 12 wherein the step of generating a first said SR rule includes the step of assigning a true value to a feature associated with the corresponding UG rule. 
     
     
       23. An apparatus for converting a lexicon and a unified grammar (UG) description into a speech recognition (SR) grammar for controlling a speech recognizer, the lexicon including a plurality of UG word entries having predefined characteristics, the UG description including a plurality of UG rules, each UG rule having at least one augmentation, the UG rules defining grammatically allowable word sequences, the apparatus incorporated within a system including a central processing unit (CPU), a memory coupled to the CPU, a system bus coupled to the CPU, and an input means coupled to the system bus for receiving a digital signal representative of an acoustic signal, the apparatus comprising: a lexical processor for generating from the lexicon a plurality of SR word entries corresponding to said plurality of UG word entries; and   a sequence and rule generator for generating from the lexicon and the UG description a plurality of permissible SR word sequences and a plurality of SR rules in said SR grammar, said permissible SR word sequences and said SR rules corresponding to the augmentations of said UG rules.   
     
     
       24. The apparatus of claim 23 wherein at least one word of at least one of said permissible SR word sequences is selected from said UG word entries by matching a predetermined characteristic of said selected word to an element of one of said UG rules. 
     
     
       25. The apparatus of claim 23 wherein one said SR rule corresponds to at least two of said grammatically allowable word sequences defined by one of said UG rules. 
     
     
       26. The apparatus of claim 23 wherein said speech recognizer comprises a second plurality of program instructions and data stored in the memory. 
     
     
       27. The apparatus of claim 23 wherein said speech recognizer comprises a plurality of program instructions and data stored in a second memory of a second computer system. 
     
     
       28. The apparatus of claim 23 wherein said lexical processor further generates a plurality of derivative word forms in said SR grammar corresponding to said plurality of UG word entries. 
     
     
       29. The apparatus of claim 23 further comprising a natural language (NL) generator for generating an NL grammar for controlling an NL processor coupled to said speech recognizer. 
     
     
       30. The apparatus of claim 29 wherein said speech recognizer propagates the predefined characteristic of said selected word to the NL processor. 
     
     
       31. The apparatus of claim 23 wherein an element of one of said plurality of SR rules refers to another one of said plurality of SR rules. 
     
     
       32. The apparatus of claim 23 further comprising a parser for converting the UG description into a LISP format. 
     
     
       33. The apparatus of claim 23 wherein a feature associated with the corresponding UG rule is assigned a true value for generating one said SR rule. 
     
     
       34. A speech recognition system for recognizing a digitized acoustic signal, said system controlled by a speech recognition (SR) grammar corresponding to a lexicon and a unified grammar (UG) description, the lexicon including a plurality of UG word entries having characteristics, the UG description including a plurality of UG rules defining grammatically allowable word sequences, said system comprising: a central processing unit (CPU);   a memory coupled to the CPU;   a system bus coupled to the CPU;   an acoustic input device coupled to the system bus for receiving the digitized acoustic signal;   a lexical processor for generating from the lexicon a plurality of SR word entries corresponding to said plurality of UG word entries; and   a sequence and rule generator for generating from the lexicon and the UG description a plurality of permissible SR word sequences and a plurality of SR rules in said SR grammar, said permissible SR word sequences and said SR rules corresponding to the augmentations of said UG rules.   
     
     
       35. The apparatus of claim 34 wherein at least one word of at least one of said permissible SR word sequences is selected from said UG word entries by matching a predetermined characteristic of said selected word to an element of one of said UG rules. 
     
     
       36. The apparatus of claim 34 wherein one said single SR rule corresponds to at least two of said grammatically allowable word sequences defined by one of said UG rules. 
     
     
       37. The apparatus of claim 34 wherein said lexical processor further generates a plurality of derivative word forms in said SR grammar corresponding to said plurality of UG word entries. 
     
     
       38. The apparatus of claim 34 further comprising a natural language (NL) generator for generating an NL grammar for controlling an NL processor coupled to said speech recognizer. 
     
     
       39. The apparatus of claim 38 wherein said speech recognition system propagates the predefined characteristic of said selected word to the NL processor. 
     
     
       40. The apparatus of claim 34 wherein an element of one of said plurality of SR rules refers to another one of said plurality of SR rules. 
     
     
       41. The apparatus of claim 34 further comprising a parser for converting the UG description into a LISP format. 
     
     
       42. The apparatus of claim 34 wherein a feature associated with the corresponding UG rule is assigned a true value for generating one said SR rule. 
     
     
       43. The method of claim 18 further comprising the step of providing feedback from said NL processor to said speech recognizer. 
     
     
       44. The method of claim 43 further comprising the step of providing a second validated word sequence to said NL processor. 
     
     
       45. The apparatus of claim 29 further comprising a feedback loop between said NL processor and said speech recognizer for causing said speech recognizer to provide a second validated word sequence.